Alicante, Granada and La Coruña to compete as seat for the Spanish AI Agency

In early October, the Spanish government published the criteria for selection of the seat for the Spanish Agency for the Supervision of Artificial Intelligence (AESIA). Spain will thus become the first country in the European Union (EU) to set up a supervisory agency for artificial intelligence (AI).

As its name implies, AESIA will supervise the creation, use and commercialisation of AI systems, especially those that might pose a threat to public safety or affect fundamental rights (such as the right to privacy). AESIA will also engage in training, dissemination and awareness activities for a responsible, sustainable and reliable use of AI systems.

Following the plan to decentralise state functions, the selection will be carried out by the Spanish Ministry of Territorial Policy. Decision criteria include the existence of research activities in the area of AI, which would involve top-notch researchers based in the region, and the presence in the region of numerous companies active in the information and communications technology (ICT) sector.

Several cities, including Alicante, Granada or La Coruña are planning to respond to this call and to apply to become the Spanish AI capital. It is expected that AESIA will begin operations in June 2023.

This comes at a time when the EU is working towards a regulatory framework proposal for AI. As part of the “Artificial Intelligence for Europe” strategy, the European Commission presented a Proposal for Regulation, setting out harmonised rules on artificial intelligence (“AI Act”).

The AI Act introduces a classification of risks related to AI systems (also known as a “risk-based approach”). According to the proposed text, there are four levels of risks: minimal or no risk, limited risk, high risk, and unacceptable risk. An example of the latter one are social scoring AIs, which are used by certain governments outside of the EU to track the trustworthiness of their citizens.

For each level of risk, a set of measures is imposed on the companies, ranging from transparency rules to a wholescale ban on certain activities. In the case of non-compliance, companies will be subject to significant sanctions, similar to those contemplated under the General Data Protection Regulation (GDPR). In turn, each Member State will have to establish one or several national competent authorities to ensure the application and implementation of the AI Act.

Where only one competent authority exists, it will also act as a national supervisory authority. If a Member State chooses to establish several competent authorities, one of them will have to be designated as the national supervisory authority. The supervisory authority (such as the one being set up in Spain) will supervise the application and implementation of the AI Act.

The European Artificial Intelligence Board will be created on the supranational level, to collect and share best practices among Member States, as well as to contribute to their uniform application of the AI Act throughout the EU. Adoption of the AI Act is expected by mid-2023. In parallel, the European Commission has also presented a Directive on adapting non-contractual civil liability rules to artificial intelligence (“AI Liability Directive”).

None of the two legal acts regulates AI-related issues, which concern intellectual property rights, such as the authorship of AI-generated works or patentability of AI-generated inventions.